Has the AI bubble finally burst? The echoes of the dot-com crash in 2001 are growing louder - Alan Lazenby
As we navigate the turbulent waters of 2025's financial markets, one question looms large over investors: has the artificial intelligence (AI) bubble finally burst? The echoes of the dot-com crash in 2001 are growing louder, and recent developments suggest we may be on the cusp of a similar reckoning.
Having advised clients through multiple market cycles, I've seen hype drive valuations to unsustainable heights before. Today, with AI stocks dominating headlines and portfolios, it's time to examine the parallels and warning signs.
Let's take a trip down my memory lane. The dot-com era was fuelled by unbridled optimism about the internet's potential. Companies with little more than a ".com" suffix commanded astronomical valuations, often with negligible revenues or profits. Pets.com, Webvan, and countless others burned through capital on promises of future dominance, only for the bubble to pop in March 2000.
By 2002, the Nasdaq had plummeted 78 per cent, wiping out trillions in wealth. The crash wasn't just about overvaluation; it exposed fraudulent accounting, speculative mania, and a disconnect between hype and real-world utility.
Fast-forward to today, and the AI boom bears striking resemblances. The International Monetary Fund recently warned that the US AI investment surge could mirror the dot-com bust, with valuations detached from fundamentals. OpenAI's eye-watering $500bn valuation in early 2025 - despite projecting just $12.7bn in revenue this year - exemplifies this exuberance. Anthropic's valuation has nearly trebled to $170bn in mere months, while the global AI market, valued at around $244bn this year, is projected to swell to over $800bn by decade's end. Yet, as Paul Krugman noted in his recent analysis, the frenzy around AI parallels the 1990s hype, where infrastructure overinvestment preceded disaster.
But are there differences? Proponents argue AI is more grounded, with tangible applications in healthcare, automation, and data analysis. Unlike many dot-com firms, today's AI giants like Nvidia boast real revenues: Nvidia's soared past $100bn annually. However, history teaches us that even profitable companies can falter when bubbles deflate. The dot-com crash didn't erase the internet; it pruned the excesses, allowing survivors like Amazon to thrive.
More concerning are the Enron-like undercurrents in the AI ecosystem. Nvidia, the poster child of AI hardware, has been accused of vendor financing practices that artificially inflate demand. Reports reveal Nvidia extending billions in loans to suppliers and customers, such as CoreWeave, in which it holds a stake. This circular financing, where Nvidia funds purchases of its own GPUs, often used as collateral for further debt, echoes the telecom bubble's playbook, as seen with Lucent and Nortel. Startups are pledging Nvidia chips for up to $10bn in loans, creating a web of interdependencies that could unravel if AI hype cools. Wall Street analysts have flagged these as "very troubling" self-investment schemes, reminiscent of Enron's off-balance-sheet manipulations that hid risks until collapse.
Do we expect valuations to fall from here? Signs point to yes. MIT studies show 95 per cent of generative AI investments yielding minimal returns, and OpenAI's Sam Altman has cautioned of overexcitement akin to the dot-com era. With US household wealth heavily tied to tech equities and GDP growth leaning on IT spending, a correction could ripple globally. The Nasdaq's recent volatility, down 15 per cent from July peaks amid AI scepticism, hints at cracks forming.
Yet, bubbles don't burst overnight. Investors should heed lessons from 2001: diversify, scrutinise fundamentals, and avoid chasing momentum.
Alan Lazenby is managing director and IFA at Lazenby’s Financial Services.